Add framework for speech modules (asr/tts)

git-svn-id: http://svn.freeswitch.org/svn/freeswitch/trunk@457 d0543943-73ff-0310-b7d9-9358b9ac24b2

src/include/switch_core.h

@@ -732,6 +732,72 @@ SWITCH_DECLARE(switch_status) switch_core_file_seek(switch_file_handle *fh, unsi
 SWITCH_DECLARE(switch_status) switch_core_file_close(switch_file_handle *fh);
 ///\}
 
+///\defgroup speeech ASR/TTS Functions
+///\ingroup core1
+///\{
+/*! 
+  \brief Open a speech handle
+  \param sh a speech handle to use
+  \param module_name the speech module to use
+  \param flags asr/tts flags
+  \param pool the pool to use (NULL for new pool)
+  \return SWITCH_STATUS_SUCCESS if the handle is opened
+*/
+SWITCH_DECLARE(switch_status) switch_core_speech_open(switch_speech_handle *sh, char *module_name, unsigned int flags, switch_memory_pool *pool);
+
+/*!
+  \brief Feed data to the ASR module
+  \param sh the speech handle to feed
+  \param data the buffer of audio data
+  \param len the in-use size of the buffer
+  \param rate the rate of the audio (in hz)
+  \param flags flags in/out for fine tuning
+  \return SWITCH_STATUS_SUCCESS with possible new flags on success
+*/
+SWITCH_DECLARE(switch_status) switch_core_speech_feed_asr(switch_speech_handle *sh, void *data, unsigned int *len, int rate, unsigned int *flags);
+
+/*! 
+  \brief Get text back from the ASR module
+  \param sh the speech handle to read
+  \param buf the buffer to insert the text into
+  \param buflen the max size of the buffer
+  \param flags flags in/out for fine tuning
+  \return SWITCH_STATUS_SUCCESS with possible new flags on success
+*/
+SWITCH_DECLARE(switch_status) switch_core_speech_interpret_asr(switch_speech_handle *sh, char *buf, unsigned int buflen, unsigned int *flags);
+
+/*! 
+  \brief Feed text to the TTS module
+  \param sh the speech handle to feed
+  \param text the buffer to write
+  \param flags flags in/out for fine tuning
+  \return SWITCH_STATUS_SUCCESS with len adjusted to the bytes written if successful
+*/
+SWITCH_DECLARE(switch_status) switch_core_speech_feed_tts(switch_speech_handle *sh, char *text, unsigned int *flags);
+
+/*! 
+  \brief Read rendered audio from the TTS module
+  \param sh the speech handle to read
+  \param data the buffer to read to
+  \param datalen the max size / written size of the data
+  \param rate the rate of the read audio
+  \param flags flags in/out for fine tuning
+  \return SWITCH_STATUS_SUCCESS with len adjusted to the bytes written if successful
+*/
+	SWITCH_DECLARE(switch_status) switch_core_speech_read_tts(switch_speech_handle *sh, 
+														  void *data,
+														  unsigned int *datalen,
+														  unsigned int *rate,
+														  unsigned int *flags);
+/*! 
+  \brief Close an open speech handle
+  \param sh the speech handle to close
+  \param flags flags in/out for fine tuning
+  \return SWITCH_STATUS_SUCCESS if the file handle was closed
+*/
+SWITCH_DECLARE(switch_status) switch_core_speech_close(switch_speech_handle *sh, unsigned int *flags);
+///\}
+
 ///\defgroup misc Misc
 ///\ingroup core1
 ///\{

src/include/switch_loadable_module.h

@@ -71,6 +71,8 @@ struct switch_loadable_module_interface {
 	const switch_api_interface *api_interface;
 	/*! the table of file formats the module has implmented */
 	const switch_file_interface *file_interface;
+	/*! the table of speech interfaces the module has implmented */
+	const switch_speech_interface *speech_interface;
 };
 
 /*!
@@ -133,6 +135,14 @@ SWITCH_DECLARE(switch_api_interface *) switch_loadable_module_get_api_interface(
  */
 SWITCH_DECLARE(switch_file_interface *) switch_loadable_module_get_file_interface(char *name);
 
+/*!
+  \brief Retrieve the speech interface by it's registered name
+  \param name the name of the speech interface
+  \return the desired speech interface
+ */
+SWITCH_DECLARE(switch_speech_interface *) switch_loadable_module_get_speech_interface(char *name);
+
+
 /*!
   \brief Retrieve the list of loaded codecs into an array
   \param pool the memory pool to use for the hash index

src/include/switch_module_interfaces.h

@@ -280,6 +280,46 @@ struct switch_file_handle {
 };
 
 
+/*! \brief Abstract interface to a speech module */
+struct switch_speech_interface {
+	/*! the name of the interface */
+	const char *interface_name;
+	/*! function to open the speech interface */
+	switch_status (*speech_open)(switch_speech_handle *sh,
+								 unsigned int flags);
+	/*! function to close the speech interface */
+	switch_status (*speech_close)(switch_speech_handle *);
+	/*! function to feed audio to the ASR*/
+	switch_status (*speech_feed_asr)(switch_speech_handle *sh, void *data, unsigned int *len, int rate, unsigned int *flags);
+	/*! function to read text from the ASR*/
+	switch_status (*speech_interpret_asr)(switch_speech_handle *sh, char *buf, unsigned int buflen, unsigned int *flags);
+	/*! function to feed text to the TTS*/
+	switch_status (*speech_feed_tts)(switch_speech_handle *sh, char *text, unsigned int *flags);
+	/*! function to read audio from the TTS*/
+	switch_status (*speech_read_tts)(switch_speech_handle *sh,
+									 void *data,
+									 unsigned int *datalen,
+									 unsigned int *rate,
+									 unsigned int *flags);
+
+	const struct switch_speech_interface *next;
+};
+
+
+/*! an abstract representation of a asr/tts speech interface. */
+struct switch_speech_handle {
+	/*! the interface of the module that implemented the current speech interface */
+	const struct switch_speech_interface *speech_interface;
+	/*! flags to control behaviour */
+	unsigned int flags;
+
+	/*! the handle's memory pool */
+	switch_memory_pool *memory_pool;
+	/*! private data for the format module to store handle specific info */
+	void *private;
+};
+
+
 /* nobody has more setting than speex so we will let them set the standard */
 /*! \brief Various codec settings (currently only relevant to speex) */
 struct switch_codec_settings {

src/include/switch_types.h

@@ -49,12 +49,15 @@ extern "C" {
   \enum switch_ivr_option_t
   \brief Possible options related to ivr functions
 <pre>
-    SWITCH_IVR_OPTION_SYNC  - synchronous (do everyting in the forground)
-
+    SWITCH_IVR_OPTION_NONE  - nothing whatsoever
+    SWITCH_IVR_OPTION_ASYNC - Asynchronous (do things in the background when applicable)
+	SWITCH_IVR_OPTION_FILE  - string argument implies a filename
 </pre>
  */
 typedef enum {
-	SWITCH_IVR_OPTION_SYNC = (1 << 0)
+	SWITCH_IVR_OPTION_NONE = 0,
+	SWITCH_IVR_OPTION_ASYNC = (1 << 0),
+	SWITCH_IVR_OPTION_FILE = (1 << 1)
 } switch_ivr_option_t;
 	
 /*!
@@ -215,6 +218,27 @@ typedef enum {
 
 } switch_codec_flag;
 
+
+/*!
+  \enum switch_speech_flag
+  \brief Speech related flags
+<pre>
+SWITCH_SPEECH_FLAG_TTS =			(1 <<  0) - Interface can/should convert text to speech.
+SWITCH_SPEECH_FLAG_ASR =			(1 <<  1) - Interface can/should convert audio to text.
+SWITCH_SPEECH_FLAG_HASTEXT =		(1 <<  2) - Interface is has text to read.
+SWITCH_SPEECH_FLAG_PEEK =			(1 <<  3) - Read data but do not erase it.
+SWITCH_SPEECH_FLAG_FREE_POOL =		(1 <<  4) - Free interface's pool on destruction.
+</pre>
+*/
+typedef enum {
+	SWITCH_SPEECH_FLAG_TTS =			(1 <<  0),
+	SWITCH_SPEECH_FLAG_ASR =			(1 <<  1),
+	SWITCH_SPEECH_FLAG_HASTEXT =		(1 <<  2),
+	SWITCH_SPEECH_FLAG_PEEK =			(1 <<  3),
+	SWITCH_SPEECH_FLAG_FREE_POOL =		(1 <<  4),
+
+} switch_speech_flag;
+
 /*!
   \enum switch_codec_type
   \brief Codec types
@@ -352,6 +376,8 @@ typedef struct switch_io_event_hooks switch_io_event_hooks;
 typedef struct switch_buffer switch_buffer;
 typedef struct switch_codec_settings switch_codec_settings;
 typedef struct switch_config switch_config;
+typedef struct switch_speech_handle switch_speech_handle;
+typedef struct switch_speech_interface switch_speech_interface;
 typedef void (*switch_application_function)(switch_core_session *, char *);
 typedef void (*switch_event_callback_t)(switch_event *);
 typedef switch_caller_extension *(*switch_dialplan_hunt_function)(switch_core_session *);